The GaSb QD was grown on n-type GaAs(001) wafers using MBE systems. First, to reduce the native defects on the GaAs substrate, we grew sequentially a 200-nm-thick GaAs buffer, a short-period superlattice (SPS) structure with two alternate layers (5 nm-thick Al0.3Ga0.7As and a 5-nm thick GaAs), and a 75-nm-thick GaAs layer at the growth temperature of 580°C.13–15 Then, we cooled down the substrate to 350°C with valve of arsenic (As) cracker cell and main shutter (MS) open to prevent from vaporizing the As of buffer layer. the As After that, we closed the valve of As cracker cell and the MS because the GaAs substrate is thermally stable under 350°C, as shown in Figure S1.16–18 Next, a gallium (Ga) droplet was formed on the top of the GaAs layer in an ultra-high vacuum of less than 1 × 10− 9 Torr at 250°C for 10.6 s, as shown in Figure S2. The time was equivalent to the growth time of the 5-monolayer (ML) GaAs. After cooling the substrate to 180°C, the Antimony(Sb) dimer was supplied to the Ga droplets at 180°C for 5 min, resulting in the diffusion of Sb to the Ga droplets, as shown in Fig. 1(a). Thus, the GaSb QD was grown by supplying the Sb to the Ga droplets on the GaAs surface, as shown in Fig. 1(b). For the QD without a capping layer, the growth process was complete. A two-step process was used to prevent the QD from melting while the GaAs capping layer was being grown. First, the GaAs was grown on top of the surface at 280°C to prevent the GaSb QD from melting without growth stop. Next, the substrate temperature was increased to 500°C for the growth of an additional GaAs capping layer under an optimized condition. As a result, a ~ 17-nm-thick GaAs capping layer was grown. The main parameters of the growth condition, which consisted of the substrate temperature (Ts), beam equivalent flux (FAs), and growth time (tg), were optimized, as shown in Figure S1.
To confirm the formation of the QD, we measured the structural and morphological characteristics of the sample without the GaAs capping layer using atomic force microscopy (AFM), scanning electron microscope (SEM), and transmission electron microscope (TEM). First, the surface morphology of the sample was analyzed using AFM to confirm the shape of the QD. The AFM data in Fig. 2(a) clearly present the morphological shape of the film surface containing the QD and the wetting-like layer (WLL), with the root mean square roughness (Rq) being ~ 3 nm. Second, SEM and TEM were used to confirm the shape, distribution density, and diameter of the QDs. From the SEM and TEM data shown in Fig. 2(b) and (c), an average diameter of 100 +/- 5 nm was obtained for the QD. Focusing on the zone axis at the [111] direction, it is found that the nanostructure of the GaSb QD is not spherical nor pyramidal-shape; rather, the QD had a nanowire-like pillar shape with zinc-blende structure. The density of the QDs was less than 10 per µm2, as shown in Fig. 2(b) and (c). Furthermore, considering Fig. 2(c) and Figure S2, the GaSb nanostructure was formed via the self-catalyst nanowire formation mechanism,5 rather than the SK-growth mechanism.17–20 Specifically, the Ga droplet was formed during 10.6 s, which is the same as deposition time of the 5-ML GaAs, as shown in Figure S2. After that, a 40nm-thick Sb layer was deposited on the GaAs surface, as shown in Figure S3. GaSb nanowire was grown under the growth condition of the diffusion-limited process for the Sb atom. Since the migration of Sb was slow at 180°C,21 the amount of Sb diffused into the Ga droplet was not sufficient. In this case, the Ga droplet was used as a self-catalyst for the QD growth like as the nanowire growth. During the growth of the QDs, the WLL observed around the GaSb nanostructure is not attributed to the growth of the GaSb QD. Most of the Sb is considered to be vaporized during the increase in the substrate temperature up to 280°C, and the origin of the WLL is attributed to the intermixing of residual Sb into GaAs layer.
To investigate the nanostructure of the samples further, we analyzed the cross-section TEM images, as shown in Fig. 3(a); a clear difference was observed, i.e., the top and bottom areas of the image indicate the GaSb QD and the GaAs layer, respectively. Each red square from the bottom to the top shown in Fig. 3(a) were used for fast Fourier transform (FFT), which are related to Fig. 3(b) and (c), respectively. Using the FFT images in Fig. 3(b) and (c), the elongated direction of the QDs is not parallel to the [001] direction of the GaAs substrate. The cross-sectional image reveals that the QD grown along the [111] direction of the GaAs layer had a zinc-blende structure. Using the reciprocal lattice constants of the GaAs layer in each direction, which were 2.93 nm− 1, 2.96 nm− 1, and 3.41 nm− 1 from the selected-area diffraction (SAD) pattern, the diffracted plane was indexed based on the reported data, as shown in Fig. 3(b).22–24 Moreover, using the reciprocal lattice constants of GaSb in each direction, which were 2.60 nm− 1, 2.71 nm− 1, and 3.17 nm− 1, the diffracted plane was indexed, as shown in Fig. 3(c). Based on the reported lattice constants of (111), (11 − 1), and (002), we indexed the plane of the reciprocal lattice following the length of each direction, as shown in Fig. 3(c). Furthermore, comparing with reported lattice constant of GaSb,25–26 we confirmed that difference of lattice constant in GaSb QD is ~ 1.8%. From this result, we suggest that due to the high lattice mismatch between GaAs and GaSb over 7%, growth direction of GaSb QD was changed to reduce the strain, resulting in the most stable growth of the QD with minimum interfacial strain. In this process, the strain was reduced by ~ 1.8% along the [111] direction. As a result, we can confirm that the QD is well grown along [111] direction of GaAs, and it has ~ 1.8% lattice strain. As a result, the GaSb QD was grown by nanowire formation mechanism with limited amount of group III material.
To investigate the band alignment between the GaSb QD and GaAs layer, we performed a PL measurement. Figure 4 (a) presents the PL spectra for a QD sample measured at 16 K by changing the laser power from 1 mW to 30 mW. Considering the peak separation using Gaussian fitting, the PL peak at 1.05 eV can be attributed to the GaSb QDs, where the full width at half maximum (FWHM) is approximately 258 meV. The energy of the PL peak emission is somewhat lower than the GaSb QDs.27–30 Three peaks are observed at 1.47 eV, 1.24 eV, and 1.32 eV; the first peak is attributed to bulk GaAs, and the other two peaks are attributed to a GaAsxSb1−x WLL. As the laser power density increases, the peak at 1.05 eV from the GaSb QDs is observed to blue-shift up to 1.13 eV. However, the blue shift of the other two peaks at 1.24 and 1.32 eV from GaAsxSb1−x WLL is weakened, as shown in Fig. 4(b). The behavior of clear blue shift with an increasing laser power at 1.05 eV is caused by the QD structure, because the shift originates from the Coulomb interaction and QD state-filling of the holes due to the spatial separation of holes in the GaSb QD and the attracted electrons confined in the nearby GaAs regions.31–33 On the contrary to the behavior, the peak shifts at 1.24 eV and 1.32 eV show the different from the nature of QDs. In addition, to confirm the GaSb QD properties, we investigated temperature-dependent PL, as shown in Fig. 4(c). Some properties were related to the QD structure of the GaSb embedded in the WLL, as shown in Fig. 4(c) and (d). In detail, as the temperature increases, an interlayer exciton is generated by the activated phonon, due to the interaction of holes in the GaSb QD with electrons in the GaAsxSb1−x WLL. This process is reflected in the energy shift of the interlayer exciton to a low energy, as shown in Fig. 4(c) and 4(d). Since the calculated exciton Bohr radius in previous reported papers, ~ 20 nm, is similar to the dimension of the GaSb QD, as shown in Fig. 2,34 a weak quantum confinement effect was generated in the QD. This effect contributed to the increase in peak intensity associated with the GaSb QD at 300 K. In addition, under the contribution of the phonon to the interlayer exciton, the PL intensity of the GaAs substrate and GaSb WLL is reduced. These optical characteristics indicate that a type-II band alignment was well formed in the GaSb QD/GaAs hetero-structure.
To confirm the type-II band alignment of the GaSb QD/GaAs hetero-structure, we investigated the carrier transfer mechanism by performing TRPL measurements at 300 K. Figure 5(a) presents the decay profile of the PL at 1.29 eV (WLL) and 0.99 eV (QD). The decay curve at 0.99 eV indicates mono-exponential behavior, which can be ascribed to the typical type-II staggered band alignment of GaSb QDs, and the long decay time is attributed to the reduced spatial overlap between the electrons in the WLL and the holes in the QDs. In comparison with long decay time, the decay curve of 1.29 eV presents a two-step process composed of a faster initial and slower tail component, as shown in the fitting parameters in Table 1. The average decay times < t > of the QDs were estimated as a linear sum of weighted multiple exponentials, where Ai=1,2,… refers to the weighting coefficient for each exponential, and τi=1,2,… indicates the corresponding fitted decay characteristic times.35 In our fittings, we used up to I = 2, which provides a reasonable fit for the measured values. The average decay time varied from 2.3 to 12.5 ns from the decay curves at the different peak position as shown in Fig. 5(b). From decay times, we can estimate the charge-transfer rate constant kct following equation:
where tWLL and tTran are the average emission lifetimes of the GaSb WLL and the charge transfer, respectively. For the charge-transfer dynamics between a delocalized continuum state and a localized state, e.g., between a 2D wetting layer and a 0D QD, the functional form of this many-state Marcus model is as follows:
where ρ(E), H(E), and ΔG indicate the density of the accepting state, the electronic coupling, and the energy difference between the donor and acceptor energy levels, respectively.36-38 The density of the states, ρ(E), is obviously different between a WLL (3D) and a QD (0D). The electronic coupling, H(E), which depends on the physical overlap between the transferred electron in its initial and final states, can be independent of energy due to the physical structure. ΔG can be expressed as ΔG = Eh(QD)-Eh(WLL), where Eh(QD) and Eh(WLL) are the energies of the hole at the QD and WLL, respectively. Since the PL spectrum is strongly related to the hole energy of the WLL and QD in a type-II system, we takeΔG to be linear relation with energy. As shown in Figure 5(c), the charge-transfer rate linearly increases in overall range, while the slope in the QD region has changed drastically. This sudden change is originated from the transition of ρ(E) between the WLL and QD. As a result, we cross check that the GaSb QD is well grown on the GaAs substrate and GaSb QD has type-II band structure with GaAs.
Following the reported paper39, the PL data at 14 K were fitted with four peaks to calculate the E-IQE of the GaSb QD, as shown in Figure 6(a) and (b); the peaks corresponding to GaSb QD, WLL1, WLL2, and GaAs. However, the PL data at 300 K, fitted with one or two peaks, indicate that non-radiative emission process is increasing at a high energy, indicating that the effective charge transfer to the GaSb QD. Finally,we can extract the 15+/-0.2% E-IQE using the ratio of the GaSb QD fitting area. In conclusion, the self-grown GaSb QD on GaAs with a well-defined structure has a high E-IQE. To confirm the relationship between E-IQE and the lattice strain, we calculated the total band structure depending on the lattice strain in GaSb, as shown in Figure S4 (a) and (b). Following the simulation data, the band gap of GaSb with a 1.8% strain along the [111] direction is smaller than that of GaSb without strain. To be specific, the valance band degeneracy is broken by the lattice strainand, and the band gap with a strain become smaller than that without the strain, as shown in Figure S4 (c) and (d); this is consistent with the red shift of the QD PL. In general, the reported results indicate that the maximum efficiency of a QD decreases within the Shockley–Queisser limit, i.e., the lattice strain reduces the efficiency.40 However, since the lattice strain from the self-grown GaSb QD in this study was minimized, we optimized the efficiency by controlling the lattice strain. In this self-grown QD formation, where the interfacial strain is minimized by changing the growth direction, the self-aligned type-II band structure with an optimized strain can affect the QD size and the formation of an effective band structure; this results in an improved efficiency compared with that of samples with a higher strain reported other paperss.41-42